I'm here testing alongside your video, with the same method of just raw vocals recorded on the spot. I entered the same prompt on both AI's, including instruments (piano and voice), style (singer songwriter) and mood (emotional, bittersweet). I made it simple on purpose, I wanted a clean result with not a lot of variables. 1) On Suno: 1.a) Extending from 00:01, 4 generations: even though all 4 songs were absolutely lovely, none of them cloned my voice at all. I guess we can rule out extending from 00:01 as a viable method. 1.b) Extended from the end of the clip, 4 generations: now it's a whole different animal. It did do a fairly good job cloning my voice and added instruments too, on all 4 generations. Audio quality not that great, as expected from Suno, but it did do a good job at voice cloning. I must admit I didn't expect this! Pleasantly surprised. I then recorded about the same vocal snippet into a .WAV file and loaded it into Udio. 2) On Udio: 2.a) Extended from 00:01 (extend and crop), 4 generations: Udio added instruments and made good songs (not as good as the ones made by Suno, and much better audio quality as we've grown to expect), but didn't clone my voice at all. Definitely ruling out the 00:01 extensions as a viable method for voice cloning, both on Udio and Suno. The AI does not read the rest of the sample for information, apparently - it just discards it entirely. 2.b) Extended from the end of the clip (extended with no cropping), 4 generations: even though it added no instruments at all on any of the 4 generations, even with manual mode on and the prompt mentioning piano (it probably needs to be prodded a lot more for that), it cloned my voice so seamlessly and with such quality I barely noticed when my clip ended and the AI portion began to kick in. 3) Conclusions: As far as just voice cloning goes, my vote goes to 2.b (Udio - extend with no cropping). If I can figure out a way to easily prompt Udio into adding instruments, it would be a clear winner. (Yup, I'll definitely spend a bunch more credits trying to prod it into cooperation before I give up. I'll try with manual mode off, as well.) Now, if the voice cloning you're aiming for doesn't need to be that high quality, but you definitely need a whole song built around it, my vote goes to 1.b (Suno - extend from the end of the clip). It's a good compromise between voice and audio quality and easy results that are actually usable from the get-go.
I tested the same and I can confirm your conclusions. Suno somehow polishes up my voice. Nice but not exactly a clone. Having a hard time to add instruments to my a capella recording in Udio. Any advise is welcome.
@@SpotiDJ I haven't done further tests, but I suppose (maybe this is just my wishful thinking) Udio can eventually add instruments even when you're providing an a capella sample with enough persistence and different prompts (manual or not). Personally, I don't see why it would never add any instruments, even when prompted clearly to do so. But I could be wrong, of course. I should run about 20 more generations with different settings before I throw the towel on this!
Funny that you mention loving our songs. I’ve created over a 1,000 songs with Suno, but I’ve only posted 28 on my channel I thought were amazing. I fell in love with them, but most people are like “eh” about them. That’s ok, because it’s mostly about me getting the lyrics out of my head. I will say, that it is absolutely amazing when someone says they love one of my songs and shares how it made them feel. It’s cool to know my lyrics had a positive effect on them.
i felt that too. Few days ago, someone commented on one of my videos says how he moved with the song and shared his broken heart story. It's a great feeling.
I use Udio to make samples for editing into a song intro before feeding it to Suno. Suno seems better at understanding melody and core features in uploaded samples in exchange for overall audio quality.
@@kead_davidson It's possible that a third relay, back into Udio, could yield benefits. In my experience, the key is to make sure those 2 mins Udio "understands" for remixing are best melodical parts, so some trimming of Suno output may be needed first. Likewise for Suno, I do some basic trimming before feeding it that 1 min, so it can "extend" after having the most relevant understanding (best moments packed in that 1 min).
I cant use Suno as every single thing it generates sounds robotic and the audio quality is terrible. Both packages are irritating to use but I get better results from Udio even though they've changed the algorithm, its not quite as good as it was a few months ago.
I generated more than 500 songs using Suno. I generated Spanish , Portuguese, Englisn, Gaelic ,Hebrew all Arabic dialects, Turkish, Kurdish, Persian ,and the list is endless Though I have access to only 5 generations per day It's such fun for a retired musician 1m like me. I once tried Udio but these 2 minutes generation limit is awful at least Suno gives you a 4 minutes generation if you carefully use the prompts AI is improving in all fields just hope I got enough time to see the changes.great video by the way keep posting
After using Suno and Udio, they deffinitely each have their strengths. Udio has more tools and better audio quality, but Suno seams to be easier to direct and get the sound you want. Udio doesnt recognize prompts the same way, so now I have to relearn how to prompt Udio after spending so much time in Suno
SUNO has a 350 character limit on the prompt, UDIO has no token limit on the prompt.. I've stuffed pages that Claude, ChatGPT and Gemeini have given me when requesting prompts targeting some particular song I wanted to replicate, like now I'm trying to replicate ZZTop's TUSH. But most of the gens its making are in 4/4 non-swing.. But I think I will get it eventually. And believe it or not, the prompts I give it from those LLMs, actually work.. Someone on reddit UDIO songwriter group just published a list of about 10,000 words that UDIO recognizes in its prompt, and I had claude categorize them for me, and it included stuff like the name of albums, years, band names, there were about 12 categories.. This means the model was trained not only on the music and sounds of commercial artists but the meta data.. So if you tell UDIO to create some 1979 Disco or some 1977 Disco, they will likely have a slight difference. You can also say things like "give me a wide stereo flanger effect"..
A trick I have found with Udio is to first have a few seconds of voice then few seconds of silence. If that does not work then add a second or less of instruments at end, so that it is forced to make music. Edit: Played around with your voice a bit myself and tried those tricks. Was able to even make a duet between your regular voice and "tiny tim". You have to do a few generations to get something decent though.
Despite your technical issues on Udio, I honestly thought it did a much better job of cloning your voice--and without the overly Autotuned sound that Suno gave. (edit: Also, thank you for doing this! Very helpful.)
I've noticed that I'm starting to get quite a bit of gibberish with Udio, too. Not quite so often with Suno, but it happens. Personally, if the quality of Suno improved to that of Udio, I'd probably only use Suno.
The icons in UDIO that are given a new generation are fingerprints, if you observe them closely in the list and what is in the player, you can tell which generation is currently in your player using the finger prints. The other way to track the generations is to look at the "version" number, V1.2.1.1.2 means the first choice that was extended and of the extended the second segment generated, and so on. You can navigate it visually if you turn on the tree mode at the top of the listing on the right. Also if you want to reduce the complexity of that listing, you can name your first generation (your creation or remix) , then from that point on all the offspring will have the same name, and on top of that you can use the search bar above to filter out all other songss. The same technique can be used to find songs you've made in the past and extend or remix them. With some practice you can organize your browser in such a way as to work on multiple songs at once using a combination of the tree, search or just fingerprints or versioning, or all four, to work on multiple songs at once.. I've worked on 5 songs at a time using this.
Udio is far more faithful here. It is actually building from the audio file you supplied. Since the audio file was awful, the extension was awful. But it sounded more like what you uploaded than what Suno did.
It doesn't matter what style of music I try to create in Suno, everything has that irritating modern pop sound which I detest and the audio quality is worse than a low bitrate mp3, its quick to use and sometime generates half decent stuff if you want electronic music but they need to do some serious work to improve it.
i really hope that we see an open source model and I'll cross my fingers that it's as impressive as when the ex-Stability guys went from co-creating Stable Diffusion to Flux. Suno and Udio updates always improve one area and make another worse. I am a little obsessed with all of this and I can't see that changing (why would it? My love of music never died), but it's hard to feel in control right now or generate with consistency.
Oh another thing, since the engine has the tendency to name segments the same, if you use the search bar, it will remind you of stuff you've done in the past that went by say "Neon Rhythm" and you will find stuff you left behind that was really cool, and it will cause you to temporarily extend that stuff .. Or you can throw it into a playlist and do it later.. Oh something to notice, the track selector radio buttons that go up the left side of the sidebar track listing , you can use that to select tracks you either want to delete or put into a playlist. Eventually they will add more things to do with that feature.. But I seem to recall it was only used for deletion for a while..
Yes, you have to reload the page or something. I had a couple of those errors, today. Just got to make sure to copy the lyrics (or, I did) and make sure the settings haven't reverted.
I used Suno and it sometimes I definitely hear my voice and sometimes I can’t recognize anything.. i wish it was more consistent.. if they become more consistent with using our voice and not switching our voice to another voice filter that Suno uses it would be number 1
11:55 try this, do a remix of your extension, and put the remix window over your lyrics, and set the remix strength to 50% or 30%.. I'm pretty sure the problem is in the strength settings.. Cause I know what it does if the prompt is not strong enough.. IT will try to match the audio and pick a genre close to the singing, which may end up being a cappella.. What some people do is layer it over the music they want to replicate.. And upload that..
It's not broken. Remix doesn't create anything new from blank audio. There has to be something there. One thing I've found that melts pretty well is singing in the shower. It gives you some nice background noise for the computer to latch onto. You can also use partial instrumentals, and have the system just fix them. As long as there's sound there for the entire 2:11, you're in good shape.
on one of my very first generations in attempts of making a genuine song with UDIO (after only used SUNO) I managed to get a song I really liked EVEN though it actually did mess it up by giving me a female counter vocalist but it worked to the song and lyrics so I was forgiving. I've of course generated plenty of great snippets after that but never to a full song or at least me giving up on trying. My last time was the moment I decided not to pay for UDIO again until they make some proper changes. Ironically enough they've just added the 2.5 model then too so one would've think they'd be on their prime quality. On a prompt for one generation which you know give 2 alternatives I got one where he talked in a dark mystical way over some soundscapes and on the 2nd I literally got Metalcore with honestly the best fry screams I've heard so far and believe me I've tried creating something to that both before and after without luck. Every single generation basically pulled my wedgies I. I do feel kinda robbed by them when I give them money for a service with all these impressive customationable freedom but no matter my effort it just push out random inconsistency. btw gotta say you are hilarious my man! Thank you for being you XD
9:42 it will limit your lyrics if you are making a 32s segment, if you use the 130s segment it won't limit the length of the lyrics. In either case, you can add the lyrics to the prompt, and UDIO will use the hint to determine the lyrics when its doing auto-generate. So there is two ways to have UDIO sing your lyrics, using the lyrics section and/or you can stuff the lyrics into the prompt after specifying the genres.. I don't usually use Manual mode, but I'm pretty sure it will work that way in manual.. The major difference I see with MAnual mode is it not forgiving if the content going in is copyrighted.. If you use Auto mode, it won't fault you for anything in the prompt. And if a band's name or singer is mentioned , it will substitute them with a set of genres they are classified under.. There is a small chance you just might get them in your generation too. Try this put "Crosby, Stills and Nash" in the prompt, and nothing more.. Then generate like 8 130s segments.. Eventually one of them will have a group of guys singing in a CSN harmony.. Just don't expect them to sing something like "Our House" or "Just a song before I go" or "Markesh Express", it will sound often like they are trying to sing "southern cross" or "free and easy". The reason many people are on UDIO, is cause it has the ghosts of dead singers in it. Its the spookier AI audio generator.
Ya you can only extend 32 seconds at a time.. The 130s segments are actually designed to create a self contained song.. ITs the make song button practically. In fact if you go back out and go to the top level menu, and look at the create song (line at the top), it gives you the option to make a 2 minute song if you are a member. I think eventually UDIO will have a three layer interface, that will permit a terraced learning curve.
In udio is very important to set correctly the prompt and lyrics relevance, otherwise it will refuse to sing. Lyrics must be set high (50% or more) when there is full of time and space to sing. But if the quest is hard, like in a short inpaint, or when there is loud instrumental around, is better to set low (20~30%) otherwise it will refuse to sing. Is a bit counter-intuitive and may be mastered only by experience.
I have just experimented with udio after getting a reply to a comment I got, so I got only vocals at first, then prompted for distorted dirty gritty guitar ect, but tomorrow I will record my own guitar riff, then record my voice, and add it to 1 file and add that to try influence it with my sound.
Good stuff, 💯💯🤙 Bob’s a funny guy 😂>.. Udio voices and music sounds much cleaner. Suno everything has a robot sound to it, I suspect it’s heavy auto tune I’m a Udio Pro User - much better results
So, have you got a chance to play with the meta-tags (the prompts that we can provide within lyrics in [ ] like [chorus], [verse], etc) to see how can we prompt Suno to sing a verse in male voice and another in female voice?
Watching the video, I noticed that Udio tries to clone the given voice as faithfully as possible. If you sing very badly, Udio does the same. In the case of Suno, it's a different approach: it's pretty close to the timbre and intonations of the voice, but it clearly enhances the way she sings with effects, autotune and that sort of thing. That's what I think and that's my opinion, nothing more. And thank you Bob for showing us the success & fails of all your attempts.
Love your channel. Off topic question (but still relevant): are there any programs or softwares that describe the genre of a specific music composition or a song? Like if I want to make my own song using Suno or Udio and I want to mimic a well known song/melody, but I'm not really sure about the genre(s) this song is in, can a program help me determine the genre(s)? Currently what I do is tell Chatgpt to describe genres of a specific song that's well known and it does, but if a song is not known, then Chatgpt won't know it. Thanks in advfance!!
I love Udio, but I cannot get this to work. I've tried with a vocal track acapella, and with a piano track with no other instruments. No amount of tinkering has gotten me a track that has (other) instruments added. Frustrating. If you figure out any tricks for getting it to work, that could be a great video on its own! Thank you again for this video
in one word great! both tenses you filled in .01 is this correct? I am so enthusiastic, can you please explain it to me properly? Otherwise I might make mistakes. I am not very good at English, but if you write it down I can have it translated.
Just curious… how often do you use Sonu’s “report” button? The number of times I’ve gotten a generation that didn’t follow the lyrics or didn’t make any sense is pretty substantial, so I always hit it for one reason or another. Often times I’ll get a 4-minute song that doesn’t have a proper ending, so I’ll report for that. It makes no sense to start repeating the bridge or random lines from Verse 1 at 3:55 when it should have automatically given me an outtro. Of course they’ve never replied to me about the reports or refunded my credits. But I’ve definitely wasted a lot thinking it’ll get it right next time but never does.
I would say overall Suno is better at creating better sounding music but. Udio The instruments are clearer. The vocals are definitely clearer and less processed, but I’m just overall not crazy about the music it creates. I like Suno better.
Question for you! So when you you extended from the 3 seconds and it used the original audio you uploaded.. but you liked what it made after that point and changed it.. how would you actually use the song then? Since the first few seconds are ehhh
11:19 I think what you did was you didn't set prompt strength to 100% and lyric strength to 20%.. Another thing you can try is put the lyrics in the prompt and use "auto generate".. It will steal the lyrics from the prompt.. Try it..
Yeah, I figured it was something like that. I sort of had that realization while I was recording the video and it just seemed like it act up with my entire approach to do it differently so I just kept going with it.😅
@@BobDoyleMedia These tools are def quite moody 😀 Only people with nerves of steel can deal with going through 30 crap sounds just to hit that shiny gem
Which one of Udio’s paid plans is needed the use a song to monetize a youtube video? In their FAQ it says a subscription plan is needed but which one???🤷🏻♂️
@@GeorginaJungle I have played with it. However, I’m having a problem with the way the music sounds. It just sounds too much like midi instruments and while the voice output I think is terrific, people are just going to beat the music quality up if I post a video about it right now. Granted I haven’t checked it in a week so maybe they’ve updated something.
@@BobDoyleMedia Wow, I didn't know it's even been out for a little bit, thought it just came out yesterday or something. I only just made my first few songs with it... I think it has potential is all...there's not really any other music tools outside of Suno and Udio.
Personally, if I don't try at least 20 outputs by Suno or Udio, on any given prompt, for me it feels like I didn't give it a fair shot. I slightly tweak the parameters in these 20 outputs, especially if pronounciation sucks (remix) or instrument selection sucks or something else bothers me consistently in the first few outputs.
I think you’re probably right that most of us are far too impatient with these tools and assume it’s never going to get what we want after just a few tries. I know much better than that from my work in the AI art and video world though.
Bob..you must be a hoot at a Karaoke Party. Hey, do you get into MidJourney at all. If so, MidJourney is suppose to add an editor to their tools this week. When I think Editor I think photoshop but I doubt it will be that extensive.
Please make the best vocal cloning tutorial, greetings from Indonesia I am waiting for it as soon as possible 💙💙💙 Please make the best vocal cloning tutorial, greetings from Indonesia I am waiting for it as soon as possible 💙💙💙 Please make the best vocal cloning tutorial, greetings from Indonesia I am waiting for it as soon as possible 💙💙💙
How the heck do you get UDIO to give you a video of a singer? I have never got the video feature to work for me, except for one with scrolling lyrics over a picture.
Let me clarify! Those videos are produced using tools like Pika just to supplement our video so it's not just me staring at words on a screen. I do understand it can be confusing, so we're going to start making it clear with on-screen notes what we're using for those - we do get asked a LOT! Thanks for watching, and for the feedback!
I can tell you the answer to that, neither, they replicate the voice of those its familiar with, and your voice will only be sustained if it is representative of some voice in its training.. That's why when you try to sustain anything, a sound, or a beat, or some style of music, the AI has its own idea about how thats supposed to sound, its that its familiar with what it was trained on, it doesn't know you..
You should have picked a genre other than bluegrass something totally unrelated to bluegrass to see if the twang showed up. Why would you pick a genre that naturally has a twang vocal to it as part of the synthesis to test whether it is going to synthesize your twangy voice?
DO THIS FOR YOUR NEXT SHOW, UPLOAD some ragtime music from some 7 inch record that is off-copyright.. The Internet Archive has tons of them.. Also someone needs to contact TOTO about their song "Animal" that they are deeply ashamed of, go search youtube for why.. They should put it in open source, cause people could extend from it.. Well that doesn't make any sense.. I mean people could use that song to sound like TOTO.. Actually it won't cause the singer they had for that, they kicked out of the band, he sounds nothing like the rest of their singers. Anyhow, sooner or later someone is going to do it, they are going to extend a commercial artist, and that artist needn't be still employed. We could get someone who is in assisted living, and wants to still make more music with their voice.. Say, Peter Wolf of the J Geils Band.. HE could do this.. Someone just needs to go to his facility and give him a mic .. Then we could have some kid help him out with UDIO, and in no time at all Peter Wolf music.. Lights Out, Flash Flash Flash... Lights out, Flash flash flash..
While the audio quality may have improved, the composition quality has noticeably declined. Something significant has changed in the arrangements, making them worse. It also seems that certain distinctive voices have been removed, likely due to pressure from the music industry. Having worked on several projects, it’s clear to me that while the sound quality is better, the overall artistry has suffered-a real disappointment.
For the most part, the vocals are terrible on both. If you're a serious music fan. However, you get lucky sometimes, on both. Currently, I think Udio is more difficult to work with your own tracks. I like the idea of using them together, but I've mostly leaned toward Suno. I wish both had more options. Better every day.
Yeah. I've been trying to use Audimee to help me finally record demo's of songs I've written (since I'm not a singer)...and majority of the time my original vocals actually sounds better that the model. Wich says a lot about how bad they are lol. I don't know if it's the timbre of my specific voice or what, but the model usually comes out "breath-ey" sounding. I got lucky and it sounded halfway decent once so far. These examples actually sound better than what get from Audimee, but I'm not wanting to have any of the music or lyrics generated...just transform the sound of the vocal I'm feeding into it. I don't think audio or suno are for doing that. Or can they?
I don't think Suno is trying to "clone" your voice. Is trying to use the rhythm. I don't know about Udio but I never got good results with it, I think is all hype when they say is better then Suno.
@@edomar2r I do understand it’s not the purpose of the platform to clone the voice, but it does seem to borrow from it heavily and sometimes it’s dead on so it’s something I’m asked about a lot.
Udio has way better audio quality and variety in it's generations. Suno always has an artificial-ness to it that is impossible to ignore once you learn to hear it and I feel like it struggles with lesser-known genres. The upside of Suno is that it's quick and easy, while Udio has a strong learning curve. But Udio absolutely sounds better than Suno. There's no way you can earnestly try both and come away with the conclusion that Suno produces better audio.
with suno I uploaded my audio and sometimes it clones my voice, somethimes it resemble it and other times keeps the melody (and make variations) but changes the voice
@@johnstarfire I have the same experience. i find the following tends to give more accurate results: do NOT put a genre in the description, and try it without lyrics (or use lyrics different to the upload audio).
@@mbirdy924 sometimes, but randomly for what I experienced, it is very capable of mimicking my voice, but it's not predicable when he will be faithful or not, I think they could make something about this in the future to control it. I tried also Udio and also randomly gave me great similarity. I think that when we will have more control these tools will substitute or integrate in daws. What I think could be very useful could be: "make a funky bass line in C major" and he makes it, then you got this stem and then say add drums and so on, being able to mix the stems as you want. They got to focus on this and focus to substitute DAWS, instead of creating random tunes with standard lyrics that do not carry anywhere if not to train the models.
I'm here testing alongside your video, with the same method of just raw vocals recorded on the spot. I entered the same prompt on both AI's, including instruments (piano and voice), style (singer songwriter) and mood (emotional, bittersweet). I made it simple on purpose, I wanted a clean result with not a lot of variables.
1) On Suno:
1.a) Extending from 00:01, 4 generations: even though all 4 songs were absolutely lovely, none of them cloned my voice at all. I guess we can rule out extending from 00:01 as a viable method.
1.b) Extended from the end of the clip, 4 generations: now it's a whole different animal. It did do a fairly good job cloning my voice and added instruments too, on all 4 generations. Audio quality not that great, as expected from Suno, but it did do a good job at voice cloning. I must admit I didn't expect this! Pleasantly surprised.
I then recorded about the same vocal snippet into a .WAV file and loaded it into Udio.
2) On Udio:
2.a) Extended from 00:01 (extend and crop), 4 generations: Udio added instruments and made good songs (not as good as the ones made by Suno, and much better audio quality as we've grown to expect), but didn't clone my voice at all. Definitely ruling out the 00:01 extensions as a viable method for voice cloning, both on Udio and Suno. The AI does not read the rest of the sample for information, apparently - it just discards it entirely.
2.b) Extended from the end of the clip (extended with no cropping), 4 generations: even though it added no instruments at all on any of the 4 generations, even with manual mode on and the prompt mentioning piano (it probably needs to be prodded a lot more for that), it cloned my voice so seamlessly and with such quality I barely noticed when my clip ended and the AI portion began to kick in.
3) Conclusions:
As far as just voice cloning goes, my vote goes to 2.b (Udio - extend with no cropping). If I can figure out a way to easily prompt Udio into adding instruments, it would be a clear winner. (Yup, I'll definitely spend a bunch more credits trying to prod it into cooperation before I give up. I'll try with manual mode off, as well.)
Now, if the voice cloning you're aiming for doesn't need to be that high quality, but you definitely need a whole song built around it, my vote goes to 1.b (Suno - extend from the end of the clip). It's a good compromise between voice and audio quality and easy results that are actually usable from the get-go.
Wow, thanks for doing all this legwork!
@@BobDoyleMedia trying acoustic song when you tried to clone the voice and Suno
I tested the same and I can confirm your conclusions. Suno somehow polishes up my voice. Nice but not exactly a clone. Having a hard time to add instruments to my a capella recording in Udio. Any advise is welcome.
@@SpotiDJ I haven't done further tests, but I suppose (maybe this is just my wishful thinking) Udio can eventually add instruments even when you're providing an a capella sample with enough persistence and different prompts (manual or not). Personally, I don't see why it would never add any instruments, even when prompted clearly to do so. But I could be wrong, of course. I should run about 20 more generations with different settings before I throw the towel on this!
Funny that you mention loving our songs. I’ve created over a 1,000 songs with Suno, but I’ve only posted 28 on my channel I thought were amazing. I fell in love with them, but most people are like “eh” about them. That’s ok, because it’s mostly about me getting the lyrics out of my head. I will say, that it is absolutely amazing when someone says they love one of my songs and shares how it made them feel. It’s cool to know my lyrics had a positive effect on them.
I know exactly what you mean. :)
i felt that too. Few days ago, someone commented on one of my videos says how he moved with the song and shared his broken heart story. It's a great feeling.
Bob, it's my favourite video of yours for pure entertainment, you had me in stitches, it was hilarious, cheers mate!
@@Mr_Kweeba thanks so much! Despite the results, this was fun!
Udio is much smarter, but often the music and voice's aren't great.
So I use suno first and than enhance it by remix in udio. Works like a charm
I use Udio to make samples for editing into a song intro before feeding it to Suno. Suno seems better at understanding melody and core features in uploaded samples in exchange for overall audio quality.
@@Dutchruler Gonna have to give that a try...
@@kead_davidson It's possible that a third relay, back into Udio, could yield benefits. In my experience, the key is to make sure those 2 mins Udio "understands" for remixing are best melodical parts, so some trimming of Suno output may be needed first. Likewise for Suno, I do some basic trimming before feeding it that 1 min, so it can "extend" after having the most relevant understanding (best moments packed in that 1 min).
I cant use Suno as every single thing it generates sounds robotic and the audio quality is terrible. Both packages are irritating to use but I get better results from Udio even though they've changed the algorithm, its not quite as good as it was a few months ago.
@@maccagrabme hence the remix afterwards in udio on what Suno created. Works very well
I generated more than 500 songs using Suno. I generated Spanish , Portuguese, Englisn, Gaelic ,Hebrew all Arabic dialects, Turkish, Kurdish, Persian ,and the list is endless Though I have access to only 5 generations per day It's such fun for a retired musician 1m like me. I once tried Udio but these 2 minutes generation limit is awful at least Suno gives you a 4 minutes generation if you carefully use the prompts AI is improving in all fields just hope I got enough time to see the changes.great video by the way keep posting
After using Suno and Udio, they deffinitely each have their strengths. Udio has more tools and better audio quality, but Suno seams to be easier to direct and get the sound you want. Udio doesnt recognize prompts the same way, so now I have to relearn how to prompt Udio after spending so much time in Suno
SUNO has a 350 character limit on the prompt, UDIO has no token limit on the prompt.. I've stuffed pages that Claude, ChatGPT and Gemeini have given me when requesting prompts targeting some particular song I wanted to replicate, like now I'm trying to replicate ZZTop's TUSH. But most of the gens its making are in 4/4 non-swing.. But I think I will get it eventually. And believe it or not, the prompts I give it from those LLMs, actually work.. Someone on reddit UDIO songwriter group just published a list of about 10,000 words that UDIO recognizes in its prompt, and I had claude categorize them for me, and it included stuff like the name of albums, years, band names, there were about 12 categories.. This means the model was trained not only on the music and sounds of commercial artists but the meta data.. So if you tell UDIO to create some 1979 Disco or some 1977 Disco, they will likely have a slight difference. You can also say things like "give me a wide stereo flanger effect"..
I think Bob is the best I like to look at his channel always a lot of humor great keep up the good work
Good Job ! Congratulations from DJ Arhitect
A trick I have found with Udio is to first have a few seconds of voice then few seconds of silence. If that does not work then add a second or less of instruments at end, so that it is forced to make music.
Edit: Played around with your voice a bit myself and tried those tricks. Was able to even make a duet between your regular voice and "tiny tim". You have to do a few generations to get something decent though.
Despite your technical issues on Udio, I honestly thought it did a much better job of cloning your voice--and without the overly Autotuned sound that Suno gave. (edit: Also, thank you for doing this! Very helpful.)
Thanks!
I've noticed that I'm starting to get quite a bit of gibberish with Udio, too. Not quite so often with Suno, but it happens. Personally, if the quality of Suno improved to that of Udio, I'd probably only use Suno.
thanks again for all the great info!
The icons in UDIO that are given a new generation are fingerprints, if you observe them closely in the list and what is in the player, you can tell which generation is currently in your player using the finger prints. The other way to track the generations is to look at the "version" number, V1.2.1.1.2 means the first choice that was extended and of the extended the second segment generated, and so on. You can navigate it visually if you turn on the tree mode at the top of the listing on the right. Also if you want to reduce the complexity of that listing, you can name your first generation (your creation or remix) , then from that point on all the offspring will have the same name, and on top of that you can use the search bar above to filter out all other songss. The same technique can be used to find songs you've made in the past and extend or remix them. With some practice you can organize your browser in such a way as to work on multiple songs at once using a combination of the tree, search or just fingerprints or versioning, or all four, to work on multiple songs at once.. I've worked on 5 songs at a time using this.
Udio is far more faithful here. It is actually building from the audio file you supplied. Since the audio file was awful, the extension was awful. But it sounded more like what you uploaded than what Suno did.
It doesn't matter what style of music I try to create in Suno, everything has that irritating modern pop sound which I detest and the audio quality is worse than a low bitrate mp3, its quick to use and sometime generates half decent stuff if you want electronic music but they need to do some serious work to improve it.
nice comparision, i now know that the sound we imput will be the voice of the lirics, tvm
i really hope that we see an open source model and I'll cross my fingers that it's as impressive as when the ex-Stability guys went from co-creating Stable Diffusion to Flux. Suno and Udio updates always improve one area and make another worse. I am a little obsessed with all of this and I can't see that changing (why would it? My love of music never died), but it's hard to feel in control right now or generate with consistency.
@@woobeforethesun true, but all of this has happened in such a sliver of time when you look at it. Give it a year or two.
Oh another thing, since the engine has the tendency to name segments the same, if you use the search bar, it will remind you of stuff you've done in the past that went by say "Neon Rhythm" and you will find stuff you left behind that was really cool, and it will cause you to temporarily extend that stuff .. Or you can throw it into a playlist and do it later.. Oh something to notice, the track selector radio buttons that go up the left side of the sidebar track listing , you can use that to select tracks you either want to delete or put into a playlist. Eventually they will add more things to do with that feature.. But I seem to recall it was only used for deletion for a while..
Yes, you have to reload the page or something. I had a couple of those errors, today. Just got to make sure to copy the lyrics (or, I did) and make sure the settings haven't reverted.
I’m just super glad suno failed to clone u and instead fixed ur voice.
Was much much nicer to listen to. 😎🤩
I like suno for broad brush ideas, and then like udio for more granular tweaks. Also, Suno vocals are really samey. It bugs me.
I love the tools and tweaks you can make with Udio like inpainting and stuff. Just a lil harder to direct the vocals
Everything sounds like a robot in Suno, its pretty unusable for me as I cant stand that type of modern pop sound.
I used Suno and it sometimes I definitely hear my voice and sometimes I can’t recognize anything.. i wish it was more consistent.. if they become more consistent with using our voice and not switching our voice to another voice filter that Suno uses it would be number 1
11:55 try this, do a remix of your extension, and put the remix window over your lyrics, and set the remix strength to 50% or 30%.. I'm pretty sure the problem is in the strength settings.. Cause I know what it does if the prompt is not strong enough.. IT will try to match the audio and pick a genre close to the singing, which may end up being a cappella.. What some people do is layer it over the music they want to replicate.. And upload that..
It's not broken. Remix doesn't create anything new from blank audio. There has to be something there. One thing I've found that melts pretty well is singing in the shower. It gives you some nice background noise for the computer to latch onto. You can also use partial instrumentals, and have the system just fix them. As long as there's sound there for the entire 2:11, you're in good shape.
on one of my very first generations in attempts of making a genuine song with UDIO (after only used SUNO) I managed to get a song I really liked EVEN though it actually did mess it up by giving me a female counter vocalist but it worked to the song and lyrics so I was forgiving. I've of course generated plenty of great snippets after that but never to a full song or at least me giving up on trying.
My last time was the moment I decided not to pay for UDIO again until they make some proper changes. Ironically enough they've just added the 2.5 model then too so one would've think they'd be on their prime quality.
On a prompt for one generation which you know give 2 alternatives I got one where he talked in a dark mystical way over some soundscapes and on the 2nd I literally got Metalcore with honestly the best fry screams I've heard so far and believe me I've tried creating something to that both before and after without luck. Every single generation basically pulled my wedgies I. I do feel kinda robbed by them when I give them money for a service with all these impressive customationable freedom but no matter my effort it just push out random inconsistency.
btw gotta say you are hilarious my man! Thank you for being you XD
I started using Udio and ran into a similar error the other night. You should try the prompt, 1956, rockabilly, sun records with one of your projects
@@christopherkann9255 I absolutely will! Thanks!
9:42 it will limit your lyrics if you are making a 32s segment, if you use the 130s segment it won't limit the length of the lyrics. In either case, you can add the lyrics to the prompt, and UDIO will use the hint to determine the lyrics when its doing auto-generate. So there is two ways to have UDIO sing your lyrics, using the lyrics section and/or you can stuff the lyrics into the prompt after specifying the genres.. I don't usually use Manual mode, but I'm pretty sure it will work that way in manual.. The major difference I see with MAnual mode is it not forgiving if the content going in is copyrighted.. If you use Auto mode, it won't fault you for anything in the prompt. And if a band's name or singer is mentioned , it will substitute them with a set of genres they are classified under.. There is a small chance you just might get them in your generation too. Try this put "Crosby, Stills and Nash" in the prompt, and nothing more.. Then generate like 8 130s segments.. Eventually one of them will have a group of guys singing in a CSN harmony.. Just don't expect them to sing something like "Our House" or "Just a song before I go" or "Markesh Express", it will sound often like they are trying to sing "southern cross" or "free and easy". The reason many people are on UDIO, is cause it has the ghosts of dead singers in it. Its the spookier AI audio generator.
Ya you can only extend 32 seconds at a time.. The 130s segments are actually designed to create a self contained song.. ITs the make song button practically. In fact if you go back out and go to the top level menu, and look at the create song (line at the top), it gives you the option to make a 2 minute song if you are a member. I think eventually UDIO will have a three layer interface, that will permit a terraced learning curve.
In udio is very important to set correctly the prompt and lyrics relevance, otherwise it will refuse to sing. Lyrics must be set high (50% or more) when there is full of time and space to sing. But if the quest is hard, like in a short inpaint, or when there is loud instrumental around, is better to set low (20~30%) otherwise it will refuse to sing. Is a bit counter-intuitive and may be mastered only by experience.
I have just experimented with udio after getting a reply to a comment I got, so I got only vocals at first, then prompted for distorted dirty gritty guitar ect, but tomorrow I will record my own guitar riff, then record my voice, and add it to 1 file and add that to try influence it with my sound.
I've been having the same issues with Udio, hopefully they will improve it soon
Good stuff, 💯💯🤙
Bob’s a funny guy 😂>.. Udio voices and music sounds much cleaner. Suno everything has a robot sound to it, I suspect it’s heavy auto tune
I’m a Udio Pro User - much better results
So, have you got a chance to play with the meta-tags (the prompts that we can provide within lyrics in [ ] like [chorus], [verse], etc) to see how can we prompt Suno to sing a verse in male voice and another in female voice?
Watching the video, I noticed that Udio tries to clone the given voice as faithfully as possible. If you sing very badly, Udio does the same. In the case of Suno, it's a different approach: it's pretty close to the timbre and intonations of the voice, but it clearly enhances the way she sings with effects, autotune and that sort of thing. That's what I think and that's my opinion, nothing more.
And thank you Bob for showing us the success & fails of all your attempts.
Love your channel. Off topic question (but still relevant): are there any programs or softwares that describe the genre of a specific music composition or a song? Like if I want to make my own song using Suno or Udio and I want to mimic a well known song/melody, but I'm not really sure about the genre(s) this song is in, can a program help me determine the genre(s)? Currently what I do is tell Chatgpt to describe genres of a specific song that's well known and it does, but if a song is not known, then Chatgpt won't know it. Thanks in advfance!!
x2
@@jymcaballero5748 Thanks
I love the sound and style of Udio, but it can be extremely frustrating to work with. Suno is easier, but has that Suno sound
I love Udio, but I cannot get this to work. I've tried with a vocal track acapella, and with a piano track with no other instruments. No amount of tinkering has gotten me a track that has (other) instruments added. Frustrating. If you figure out any tricks for getting it to work, that could be a great video on its own! Thank you again for this video
My only problem with Suno is that you can't choose a voice for it to use. Is it there and I'm missing it? Each song's vocal sounds the same.
in one word great! both tenses you filled in .01 is this correct?
I am so enthusiastic, can you please explain it to me properly? Otherwise I might make mistakes. I am not very good at English, but if you write it down I can have it translated.
Just curious… how often do you use Sonu’s “report” button? The number of times I’ve gotten a generation that didn’t follow the lyrics or didn’t make any sense is pretty substantial, so I always hit it for one reason or another. Often times I’ll get a 4-minute song that doesn’t have a proper ending, so I’ll report for that. It makes no sense to start repeating the bridge or random lines from Verse 1 at 3:55 when it should have automatically given me an outtro. Of course they’ve never replied to me about the reports or refunded my credits. But I’ve definitely wasted a lot thinking it’ll get it right next time but never does.
What if you put the words you sang in the song, then use extend so it can train on your words and voice better.
I would say overall Suno is better at creating better sounding music but. Udio The instruments are clearer. The vocals are definitely clearer and less processed, but I’m just overall not crazy about the music it creates. I like Suno better.
Question for you!
So when you you extended from the 3 seconds and it used the original audio you uploaded.. but you liked what it made after that point and changed it.. how would you actually use the song then? Since the first few seconds are ehhh
11:19 I think what you did was you didn't set prompt strength to 100% and lyric strength to 20%.. Another thing you can try is put the lyrics in the prompt and use "auto generate".. It will steal the lyrics from the prompt.. Try it..
Had the silent issue with Udio too. It's because you remixed a shorter than 2min clip..so it remixed silence into silence.
Yeah, I figured it was something like that. I sort of had that realization while I was recording the video and it just seemed like it act up with my entire approach to do it differently so I just kept going with it.😅
@@BobDoyleMedia These tools are def quite moody 😀
Only people with nerves of steel can deal with going through 30 crap sounds just to hit that shiny gem
keep it up
Which one of Udio’s paid plans is needed the use a song to monetize a youtube video? In their FAQ it says a subscription plan is needed but which one???🤷🏻♂️
Hi! Bob, are you going to do a video on Mureka? It's a new music generator like Suno.
@@GeorginaJungle I have played with it. However, I’m having a problem with the way the music sounds. It just sounds too much like midi instruments and while the voice output I think is terrific, people are just going to beat the music quality up if I post a video about it right now. Granted I haven’t checked it in a week so maybe they’ve updated something.
@@BobDoyleMedia Wow, I didn't know it's even been out for a little bit, thought it just came out yesterday or something. I only just made my first few songs with it...
I think it has potential is all...there's not really any other music tools outside of Suno and Udio.
Hello Bob, I am looking to clone a voice and face swap on a video and get it to lip sync. What tools would I need? I hope this is clear. Thank you.
🎉 great! Any Idea how to deal with lyric glitches, if the vocals are fine, maybe it‘s just a buggy pronounciation
Personally, if I don't try at least 20 outputs by Suno or Udio, on any given prompt, for me it feels like I didn't give it a fair shot. I slightly tweak the parameters in these 20 outputs, especially if pronounciation sucks (remix) or instrument selection sucks or something else bothers me consistently in the first few outputs.
I think you’re probably right that most of us are far too impatient with these tools and assume it’s never going to get what we want after just a few tries. I know much better than that from my work in the AI art and video world though.
Bob..you must be a hoot at a Karaoke Party. Hey, do you get into MidJourney at all. If so, MidJourney is suppose to add an editor to their tools this week. When I think Editor I think photoshop but I doubt it will be that extensive.
This was good comedy relief. lol
Thanks. Things got silly real quick. You should see what edited out...
You have to give Suno at least 59 second sample to expect any resemblance of the original voice.
I use both, and in my view Suno makes better melodies and "productions" while Udio has the better voices.
Please make the best vocal cloning tutorial, greetings from Indonesia I am waiting for it as soon as possible 💙💙💙
Please make the best vocal cloning tutorial, greetings from Indonesia I am waiting for it as soon as possible 💙💙💙
Please make the best vocal cloning tutorial, greetings from Indonesia I am waiting for it as soon as possible 💙💙💙
ElevenLabs Music will win.
I am eagerly awaiting its availability.
In my case, when i got this Backend error, i just had to reload the page. Just login again
How the heck do you get UDIO to give you a video of a singer? I have never got the video feature to work for me, except for one with scrolling lyrics over a picture.
Let me clarify! Those videos are produced using tools like Pika just to supplement our video so it's not just me staring at words on a screen. I do understand it can be confusing, so we're going to start making it clear with on-screen notes what we're using for those - we do get asked a LOT!
Thanks for watching, and for the feedback!
I hope Suno improves the sound quality in V4, because with suno I get much better songs than with udio.
I can tell you the answer to that, neither, they replicate the voice of those its familiar with, and your voice will only be sustained if it is representative of some voice in its training.. That's why when you try to sustain anything, a sound, or a beat, or some style of music, the AI has its own idea about how thats supposed to sound, its that its familiar with what it was trained on, it doesn't know you..
You should have picked a genre other than bluegrass something totally unrelated to bluegrass to see if the twang showed up. Why would you pick a genre that naturally has a twang vocal to it as part of the synthesis to test whether it is going to synthesize your twangy voice?
@@hitmusicworldwide I pretty much addressed that directly.
DO THIS FOR YOUR NEXT SHOW, UPLOAD some ragtime music from some 7 inch record that is off-copyright.. The Internet Archive has tons of them.. Also someone needs to contact TOTO about their song "Animal" that they are deeply ashamed of, go search youtube for why.. They should put it in open source, cause people could extend from it.. Well that doesn't make any sense.. I mean people could use that song to sound like TOTO.. Actually it won't cause the singer they had for that, they kicked out of the band, he sounds nothing like the rest of their singers. Anyhow, sooner or later someone is going to do it, they are going to extend a commercial artist, and that artist needn't be still employed. We could get someone who is in assisted living, and wants to still make more music with their voice.. Say, Peter Wolf of the J Geils Band.. HE could do this.. Someone just needs to go to his facility and give him a mic .. Then we could have some kid help him out with UDIO, and in no time at all Peter Wolf music.. Lights Out, Flash Flash Flash... Lights out, Flash flash flash..
While the audio quality may have improved, the composition quality has noticeably declined. Something significant has changed in the arrangements, making them worse. It also seems that certain distinctive voices have been removed, likely due to pressure from the music industry. Having worked on several projects, it’s clear to me that while the sound quality is better, the overall artistry has suffered-a real disappointment.
For the most part, the vocals are terrible on both. If you're a serious music fan. However, you get lucky sometimes, on both. Currently, I think Udio is more difficult to work with your own tracks. I like the idea of using them together, but I've mostly leaned toward Suno. I wish both had more options. Better every day.
Suno ai is the best.The only thing sith suno is you cant edit your track.
Yeah. I've been trying to use Audimee to help me finally record demo's of songs I've written (since I'm not a singer)...and majority of the time my original vocals actually sounds better that the model. Wich says a lot about how bad they are lol. I don't know if it's the timbre of my specific voice or what, but the model usually comes out "breath-ey" sounding. I got lucky and it sounded halfway decent once so far.
These examples actually sound better than what get from Audimee, but I'm not wanting to have any of the music or lyrics generated...just transform the sound of the vocal I'm feeding into it. I don't think audio or suno are for doing that. Or can they?
You can...i use ripx daw to get stems and then you
can edit mix e mastering
But you are watching it because I posted it 🤙Now that made me click the like button, because it's true 👍
Udio is a bit pitchy dawg!
The audio video are out of sync. It hurts my brain
@@choosers5177 Where is it out of sync? I’m not getting that.
Eleven labs
Suno : No More ruclips.net/video/j7byVvu5YwI/видео.htmlsi=06Vm7XlipmKUwq86
remix will remix your upload it doesnt add anything .
I don't think Suno is trying to "clone" your voice. Is trying to use the rhythm. I don't know about Udio but I never got good results with it, I think is all hype when they say is better then Suno.
@@edomar2r I do understand it’s not the purpose of the platform to clone the voice, but it does seem to borrow from it heavily and sometimes it’s dead on so it’s something I’m asked about a lot.
Udio has way better audio quality and variety in it's generations. Suno always has an artificial-ness to it that is impossible to ignore once you learn to hear it and I feel like it struggles with lesser-known genres. The upside of Suno is that it's quick and easy, while Udio has a strong learning curve. But Udio absolutely sounds better than Suno. There's no way you can earnestly try both and come away with the conclusion that Suno produces better audio.
with suno I uploaded my audio and sometimes it clones my voice, somethimes it resemble it and other times keeps the melody (and make variations) but changes the voice
@@johnstarfire I have the same experience. i find the following tends to give more accurate results: do NOT put a genre in the description, and try it without lyrics (or use lyrics different to the upload audio).
@@mbirdy924 sometimes, but randomly for what I experienced, it is very capable of mimicking my voice, but it's not predicable when he will be faithful or not, I think they could make something about this in the future to control it. I tried also Udio and also randomly gave me great similarity. I think that when we will have more control these tools will substitute or integrate in daws. What I think could be very useful could be: "make a funky bass line in C major" and he makes it, then you got this stem and then say add drums and so on, being able to mix the stems as you want. They got to focus on this and focus to substitute DAWS, instead of creating random tunes with standard lyrics that do not carry anywhere if not to train the models.
your audio is out of sync
Udio has become worse. And is malfunctioning often.
If it wasn´t for that hiss in Suno…
That's right. Suno should update its AI. Lots of noise!
why? you're not allowed to use it anyway, as soon as you publish it through digital distribution you'll be banned. so why waste money?